TIL the nn frameworks I use each initialize conv layers differently: PyTorch - kaiming he uniform TensorFlow - xavier glorot uniform Stax (JAX) - xavier glorot normal Flax (JAX) - lecun normal Haiku (JAX) - truncated normal (scaled by 1/sqrt(fan_in))

1/3) Just read this "Reward is enough" paper: sciencedirect.com/science/articl… It's really interesting, but I think it fails to recognise that evolution often sets up sub-systems with specific goals! Example: Your kidneys are for fluid regulation (in Twitter brevity).

1/6 Iterating (i.e. repeatedly composing) a function is tricky business. You can get wild (even chaotic) behavior with simple functions like r(x) = cx(1-x) That's why it's important to choose the nonlinear activation functions very carefully in neural networks.

Industry and Academic Research in Computer Vision arxiv.org/abs/2107.04902 Not your typical #computervision research paper, but a study of the dynamic between research in the industry and academia in computer vision.

The US dropped below 100,000 COVID-19 cases per day in February. Over the past week, the odds (on Polymarket) of exceeding that threshold again before the end of the year went from 50% to 75%. polymarket.com/market/will-th…

Google Health is no longer just a single team, but a significant company-wide effort that touches many of our products. Moving forward the Google Health name will encompass all our health initiatives.

Lossless Image Compression in O(n) Time — introducing QOI, the Quite OK Image Format. 300 lines of C, single header library: phoboslab.org/log/2021/11/qo…

Every year I teach CS theory, I find that I need to apologize less and less about having so much math content. CS practice is getting more & more mathy. Now I just ask students if they want to their future job to be one where they read texts of first type, or the second type:

A recent exchange on Twitter made me realize that many may be using tensors without knowing how they are represented in memory, and how it makes some operations super fast. This is standard stuff common to many tensor libraries, including our beloved NumPy and PyTorch.

Yesterday @ #MadeByGoogle we also announced the latest on the Super Res Zoom feature on #Pixel7Pro It's a project my team's been involved in since '18. This year, our teams've made it so much more powerful. You can zoom up to 30x. Let me show you in a 🧵 x.com/madebygoogle/s…

Sign up to resist big publishing's greedy shareholders and speak out for an open internet with uncensored access to knowledge. battleforlibraries.com

Indian PhD students from IISc Bangalore, who have first-authored papers at prestigious conferences like #CVPR2025, are facing unjust denials of Canadian visas. With Shocking reasons "limited employment possibilities in India" and "purpose of visit not consistent with a temp. stay."

This makes me sad: observablehq.com/@ayhanfuat/the…

Ton's of Soviet era books, a hidden treasure! x.com/MirTitles?s=20

I've spent the past ~2 weeks trying to make a chip from scratch with no prior experience. It's been an incredible source of learning so far. Progress tracker in thread (coolest stuff at the end)👇

ishan not only does it convert back losslessly it also has the noise removed

Let Readers Read: An Open Letter to the Publishers in Hachette v. Internet Archive: - Sign the Petition! chng.it/xwLDncTmDZ via Change.org India

Distributed Shampoo has dethroned Nesterov Adam marking a new era for deep learning optimization. 👑 🤘 Non-diagonal preconditioning is here! This is the AlexNet moment for optimization for deep learning. I am extremely happy. An email from 2021.

"how do you get them so chocolatey? is it the amount of chocolate that you add?" - my mom , who has never followed a recipe without deciding some ingredient or other "sounds like a bit much" and halving its quantity

One nice thing you can do with an interactive world model, look down and see your footwear ... and if the model understands what puddles are. Genie 3 creation.